Game theoretic data association for multi-target tracking with varying number of targets

We investigate a game theoretic data association technique for multi-target tracking (MTT) with varying number of targets. The problem of target state-estimate-to-track data association has been considered. We use the SMC-PHD filter to handle the MTT aspect and obtain target state estimates. We model the interaction between target tracks as a game by considering them as players and the set of target state estimates as strategies. Utility functions for the players are defined and a regret-based learning algorithm with a forgetting factor is used to find the equilibrium of the game. Simulation results are presented to demonstrate the performance of the proposed technique.